Peter Sobot


Generalist software engineer with a passion for building highly performant systems and growing others' careers.

Languages of Choice: Python, Scala, C++, Java, Ruby, C, Objective-C, Go, JavaScript (TypeScript), Swift, SQL
Areas of Focus: Performance Optimization, Teaching, Mentorship, Audio Processing, Distributed Systems, Machine Learning
👔 Work Experience
Staff Machine Learning Engineer at Spotify in New York, New York (February 2020-Present)
  • Provided technical leadership to Spotify's Audio Intelligence (MIQ) research lab, which focuses on applying Machine Learning to audio
    • Built machine learning tooling and frameworks to increase research productivity by 40x (experiments/quarter)
    • Trained, optimized, and deployed machine learning models for music information retrieval and content identification
    • First to deploy machine learning for audio processing on mobile devices at Spotify with TensorFlow Lite
    • Designed, wrote, and released Pedalboard, a high-performance Python library for audio effects with over 3,000 stars on GitHub
    • Drove cost-optimization efforts to reduce cloud spend by millions of dollars per year
    • Helped improve open source libraries including TensorFlow, TensorFlow Datasets, Apache Beam, and hnswlib
    • Temporarily acted as interim manager for 3 engineers and 5 research scientists and didn't get scared away from people management
  • Member of Spotify's company-wide incident response team, serving as the first line of defense for severe incidents and outages
  • Re-designed and re-implemented Spotify's internal approximate nearest-neighbour search library
Staff Software Engineer at Spotify in New York, New York (July 2019-February 2020)
  • Provided technical leadership, software design, and code to the Personalization Platform team (~40 people)
    • Owned and maintained the core machine learning systems for music recommendation across all of Spotify
    • Powered features including Home, Discover Weekly, Daily Mix, Radio, Suggested Songs, and Fans Also Like
    • Made contributions to open source libraries crucial to Spotify's recommender systems, including Annoy and Scio
    • Rewrote Spotify's core recommendation models to increase training speed by 5x
  • Focused on teaching, mentorship, and education
    • Led engineer onboarding bootcamp for all new US-based engineering hires
    • Redesigned introductory engineering curriculum for all global engineering hires
    • Mentored 7 engineers, resulting in 7 promotions
  • Gave dozens of internal and external talks to other teams, companies, and conferences across three continents
Senior Backend Engineer at Spotify in New York, New York (May 2017-July 2019)
  • Designed and implemented scalable backend services in Java for music recommendation and personalization
  • Built data pipelines in Scala and Python, processing hundreds of terabytes per day to serve 200m+ Spotify users
  • Won Spotify's internal company-wide hack week two years in a row (with projects sadly too secretive to list)
  • Led the Spotify NYC Musicians' club and ran bi-monthly Friday Night Live internal employee concert series
Senior Software Engineer at PagerDuty in Toronto, Ontario (November 2016-April 2017)
  • Provided technical direction, design input, mentorship and engineering work to Incident Management team
  • Authored and deployed code contributions to 34 projects in 10 languages (primarily Ruby, Javascript, and Swift)
  • Won 9 of 12 monthly engineering hack days in 2016 with projects including:
    • #oncallselfie, native Twitter integration for PagerDuty's iOS and Android apps
    • Rich HTML Email support for hundreds of thousands of PagerDuty users
    • Internal real-time display of outgoing notifications with Go, Kafka, WebSockets & WebGL
    • PagerDuty's watchOS app for displaying on-call status and scheduling on your wrist
    • Versioning and undo for PagerDuty's on-call scheduling service
  • Fastest progression from entry-level to senior engineer in company history (21 months)
Software Engineer II at PagerDuty in Toronto, Ontario (October 2015-November 2016)
  • Core contributor for product improvements on Workflow and Incident Management: Systems teams
  • Spearheaded efforts to increase overall data throughput by 100x, enabling future customer growth
  • Mentored and onboarded 2 interns and 3 new full-time employees
Software Engineer I at PagerDuty in Toronto, Ontario (February 2015-October 2015)
  • Helped ship core product enhancements including Incident Snooze and Incident Urgencies
  • Ensured reliability and uptime of PagerDuty by participating in primary on-call rotations
  • Recorded, sang, and released the BarberDuty ringtones, PagerDuty's on-call barbershop quartet that wakes you up at night
Entrepreneur in Residence at The Working Group in Toronto, Ontario (June 2014-February 2015)
  • Directed TWG Labs, a company division focused on experiments, learning and tools
  • Managed a team of 3 developers on internet-of-things projects (Raspberry Pi, iBeacons, NodeJS and Heroku)
  • Built bleeding-edge product prototypes for national Canadian brands including Tim Hortons
Founder & Lead Developer at Appstruments in Toronto, Ontario (2011-2014)
  • Crafted viral music apps for the web, Android and iOS:
    • the Wub Machine , an automatic music remixer used to make nearly 5 million remixes
    • , an endless, automatic & beatmatched radio station
Infrastructure Software Engineering Intern at Inkling in San Francisco, California (2013)
  • Rewrote Inkling's aging deployment system
  • Created a new onboarding process for engineers
  • Replicated Inkling's entire backend infrastructure for a business-critical, Apple-style product launch
Software Engineering Intern at Capcom Game Studio Vancouver in Burnaby, British Columbia (2012)
  • Finished Capcom's internal testing infrastructure for Dead Rising 3, a launch title for the Xbox One
  • Created analytics tools using real-time gameplay data to aid game designers
  • Got my first (and only, so far) video game credit
Software Developer Co-op at The Working Group (now Deloitte) in Toronto, Ontario (2011)
  • Rails and iOS development for a variety of clients and internal projects.
Web Developer Co-op at Imagination Plus (now Human_Code) in Hamilton, Ontario (2009)
  • PHP and JavaScript development. Led development of a custom PHP-driven learning management system.
🎓 Education
Bachelor of Software Engineering (with distinction) from the University of Waterloo (2009-2014)
  • Resident of VeloCity, Waterloo's startup incubator residence
  • Built MixBox, an iPad music remixing app based on the Echo Nest's (now Spotify's) Remix API (2012-2014, team of 3)
  • Wrote JoosBox, a compiler for a large subset of Java in Scala (Winter 2014, team of 3)
  • Developed "Colour of the Web," a research project to find the average colour of the internet (Fall 2010, group of 2)
🔬 Patents
🐙 Selected Open-Source Projects
  • Pedalboard, a Python library for high-performance audio effects (including VST/AU hosting)
    • Designed and built at Spotify to provide faster data augmentation for training machine learning models
    • Open-sourced to critical acclaim from the Python audio community, gaining 3,000 stars on GitHub in four months
  • keynote-parser, a Python library for decoding (and re-encoding) Apple Keynote presentation files
  • MusicBoy, a hardware project to release an album on a custom Nintendo Game Boy® cartridge
    • Built with Toronto-based electro-pop artist Pusher, who provided music and artwork
    • Based on Super Mario Land, plus an ARM-based coprocessor in the Game Pak to play back CD-quality audio
    • Presented at !!con 2019 in New York City
  • SampleScanner, a command-line app to convert hardware synthesizers into software instruments (VSTs)
📽 Selected Talks and Presentations
🎼 Activities & Interests
Music Composition, Performance, and Production (Drums, Bass, Guitar, Piano) Electronic Music Production PhotographyTwitter • Painstakingly Emulating Obsolete Digital Synthesizers from the 90s as Software Instruments • Video Production for Concerts • Brunch • Excessive Usage of Comic Sans